FIRE, the Flexible Image Retrieval Engine, is a content-based image retrieval system that I developed in cooperation with many other people at the Human Language Technology and Pattern Recognition Group of RWTH Aachen University.

The main aim of FIRE is to investigate different image descriptors and evaluate their performance. FIRE was developed in C++ and Python and is meant to be eaily extensible.

FIRE was started during my diploma thesis and then progressively extended.

Contributors include


FIRE’s most recent version is available by SVN from its Google Code project site.

Older versions are available from my old RWTH Aachen CS department website.


I will under no circumstances answer emails related to the installation or usage of FIRE that are sent directly to me. Please always write to the google group.


A new step-by-step installation how to install FIRE on a fresh Ubuntu 10.4 installation is available here.


An online demo with photographs is available here (link to RWTH Aachen)

Another demo with medical images is available here (link to RWTH Aachen)

Normally these servers are up and running. If not, please let me know. I will restart it.

The people from the Image Understanding and Pattern Recognition Group from Kaiserslautern have extended FIRE for sub image retrieval and created a very nice demo video:


If you use FIRE within your research, I would be happy if you would cite a paper of mine. Probably the most appropriate one is

Thomas Deselaers, Daniel Keysers, Hermann Ney Features for Image Retrieval: An Experimental Comparison. Information Retrieval. 2008.

Vol. 11. Issue 2. Springer. pp. 77-107.

If you are unsure which paper to cite, feel free to contact me.